Speaker
Jacob Russell Howard
(University of Oxford (GB))
Description
One possible option for the ATLAS High-Level Trigger (HLT) upgrade for higher
LHC luminosity is to use GPU-accelerated event processing. In this talk we
discuss parallel data preparation and track finding algorithms specifically
designed to run on GPUs. We present a "client-server" solution for hybrid CPU/GPU
event reconstruction which allows for the simple and flexible integration of
the specific GPU-accelerated algorithms into existing ATLAS HLT software.
The resulting speed-up of event processing times obtained with high-luminosity
simulated data are presented and discussed.
Author
Dmitry Emeliyanov
(STFC - Science & Technology Facilities Council (GB))
Co-author
Jacob Russell Howard
(University of Oxford (GB))
